Study system
We conducted field components of our study at the Larry R. Yoder Prairie Learning Laboratory in Marion, Ohio (40° 34’N, -83°5’W). This site in central Ohio has a five to six month growing season and an average annual precipitation of 998 mm. Soils are Pewamo silty clay loams with 0 to 1 percent slopes (USDA NRCS 2023a). The Larry R. Yoder Prairie is a 47-year-old tallgrass prairie restoration of ~ 4.45 hectares, characterized by Monarda fistulosa, Ratibida pinnata, Andropogon gerardii, Schizachyrium scoparium, Echinacea purpurea, Solidago juncea, and Chamaecrista fasciculata. This restoration has been managed with triennial prescribed ground-layer fires during the past decade.
In both phase 1 (inocula training) and phase 2 (growth assay) of this work, S. scoparium and R. hirta were used as representative prairie plants. Both species are widely distributed across US grasslands (USDA NRCS 2023b). S. scoparium is a perennial, C4 grass, and R. hirta is an annual/biennial forb. Both plants are considered to have “medium” fire tolerance (scale: none, low, medium, high), meaning they can resprout, regrow, or reestablish from residual seed following fire. Seeds for S. scoparium and R. hirta were purchased from Prairie Moon Nursery (Winona, MN), started in sterile potting soil (autoclave: 2 hours, 120°C), and grown for 3 weeks.
Phase 1 – inocula training: Soil inoculum was collected from the upper 15 cm of the Larry R. Yoder Prairie in September 2022 and sieved (2 cm) to remove large roots. Sieved background soil (2 cm) was collected in Columbus, OH and combined 1:1 with sand, then autoclaved twice for 2 hours at 120°C. In October 2022, 500 mL of sterilized sand:soil mix was added to 15 cm clay pots, followed by 100 mL of prairie soil inoculum, then inoculum was covered with 200 mL of sterilized sand:soil mix. Using sterile soil at the bottom and top of the pot helps prevent contamination between pots. One S. scoparium (n = 10) or R. hirta (n = 10) plant was planted in the center of each pot. Plants were grown for 3 months in a greenhouse under extended day lighting. Plants were watered 4 times daily using a drip irrigation system with 7.6 L/hr emitters for 1 min intervals. Plants were fertilized monthly with 200 mL of 200 ppm 15-0-15 (N-P-K) water soluble fertilizer.
Experimental fires
Tallgrass prairie fuel loads described in Leis and Hinman 2014 were used to design experimental fuel loads. Briefly, average prairie fuel loads of 0.7 kg/m2 were scaled down to a 176 cm2 pot soil surface area (12.36 g of fuels per pot). Experimental fuel loads consisted of wheat straws that were arranged on the soil surface of five S. scoparium and five R. hirta pots. Prior to fire, plant aboveground biomass for each pot was clipped and removed to further standardize fuel loads. Fires were ignited using a propane hand torch along to edge of the pot rim. Each fire burned for an average of 6 min with pot fires falling within +/- 15 sec of this time, and following fire, ash was removed from to top of each pot. Ash was removed to ensure that treatment effects were due to fire effects on soil microbiota rather than combustion associated nutrient pulses and changes in soil pH. Soils from all pots were collected in separate sterile bags, then soil in each bag was homogenized and stored at 4° C for 2 weeks.
Phase 2 – growth assay: To assess fire effects on plant-soil feedbacks, S. scoparium and R. hirta plants were grown in soils trained in phase 1. Seeds of both species obtained from Prairie Moon Nursery (Winona, MN) were started identically to those in phase 1. In January 2023, 500 mL of sterilized sand:soil mix was added to 15 cm clay pots, followed by 100 mL of trained soil inocula from phase 1, then inoculum was covered with 200 mL of sterilized sand:soil mix. One S. scoparium (n = 40) or R. hirta (n = 40) plant was planted in the center of each pot and starting plant heights were recorded. This produced 80 total pots, with 10 replicates for each phase 2 species x phase 1 species x fire (burn/no burn) combination. While soil microbial community composition was not assessed prior to or following burn treatments, the use of identical fuel loads and similar burn times across all pots ensures fire effects on soil biota were as uniform as possible. Further, inoculating pots with 100 mL of soil (~ 10% pot volume) ensures that inoculum effects are due to differences in soil microbiota rather than abiotic differences (Pernilla Brinkman et al. 2010). Inoculum origin was recorded for each pot to account for background variation in phase 1 inocula, and included in all statistical analyses. Note that sterile inoculum treatments were not required in this study as comparing conspecific and heterospecific trained soil is preferable when comparing plant-soil feedbacks between multiple plant species (Pernilla Brinkman et al. 2010). Plants were grown under conditions identical to phase 1. Each month of the growing period (3), plant height and tiller number (S. scoparium), as well as longest leaf length and flower number (R. hirta) were recorded. Following 3 months of growth, plant above- and belowground biomass was harvested and dried for 3 days at 60° C, then aboveground, belowground, and flower biomass (R. hirta only) were recorded. Aboveground biomass values were considered in addition to total biomass since aboveground biomass represents the actual fuel loads for fire. Root:shoot ratios were calculated for each plant by dividing above- by belowground biomass values. Phase 2 pots were then randomly assigned to pairs within species and fire treatments (as in (Pernilla Brinkman et al. 2010)), and used to calculate the following plant-soil feedback metric (Klironomos 2002; Petermann et al. 2008):
$${Feedback}_{i}=log\left[\frac{{biomass}_{i}\left(home\right)}{{biomass}_{i}\left(away\right)}\right]$$
where biomassi (home) is plant biomass of species i in soil trained by species i, and biomassi (away) is plant biomass of species i in soil trained by species j. This feedback metric was chosen as we were specifically interested in testing how plant responses to home and away soil inocula varied between fire treatments. Separate feedback metrics were calculated for both total plant biomass and aboveground biomass. Feedback values greater than 0 denote positive feedbacks (i.e., better growth in soil trained by own species), and values less than 0 denote negative feedbacks (i.e., better growth in soil trained by their other species).
Statistical analyses
Analyses were conducted in R version 4.2.2 (R Core Team, 2022). To test fire and phase 1 plant ID effects during the phase 2 growth period we used type III multivariate analyses of variance (MANOVAs) using the base MANOVA() function and the joint_tests() function in the emmeans package (Lenth 2018). MANOVAs were used to reduce type I error that arises with multiple testing (e.g., multiple ANOVA tests). The MANOVA model included tiller/leaf lengths for plants in months 1 and 3 as response variables, in addition to fixed effect terms for fire, phase 1 species (i.e., “trainer” species), and phase 2 species. Month 2 tiller/leaf lengths were not included due to high collinearity with month 3 measurements (r < |0.8|). The MANOVA model also controlled for starting plant heights, greenhouse row, and inocula source pot. Model residuals were visually assessed for normality and met model assumptions. Following significant main effects, custom contrasts were applied testing fire effects on phase 2 plant growth in inter- and intraspecific soil treatments (e.g., S. scoparium growth with burned or unburned S. scoparium vs. R. hirta trained inocula) using the contrast() function.
To test fire and phase 1 plant ID effects on phase 2 plant biomass, a separate MANOVA that included aboveground mass, total mass, flower mass, and root:shoot ratios as response variables was used. All fixed effect and covariate terms were identical to the growth MANOVA described above. Due to high collinearity between belowground and total biomass (r < |0.8|), belowground biomass was omitted from the model. Model residuals were visually assessed for normality and met model assumptions. Then, custom contrasts similar to those described above were applied.
To test fire and plant species effects on plant-soil feedbacks we used type III analyses of variance (ANOVA) that included feedback metrics as response variables, and phase 1 plant ID and fire treatment as fixed effects, as well as their interaction term. Model residuals were assessed as above,